Deep learning does not work with the RTX50 series with CUDA version 12.8 or higher.
Posted: Mon Feb 17, 2025 12:10 pm
I recently purchased rtx5090, and since tensorflow does not support cuda version 12.8 of rtx5090 yet, I cannot do deep learning via faceswap. If tensorflow supports cuda version 12.8 or higher in the future, I wonder if faceswap deep learning training will be possible using rtx5090. I would like to know if the faceswap development team plans to support the rtx50 series in the future. And I would like to know if the latest high-performance graphics cards such as rtx5090 require an Ubuntu environment such as wsl2 rather than windows for faceswap deep learning tasks. And I would also like to know what benefits there are for supporting faceswap.